Privacy-Preserving Collaborative Learning for Mobile Health Monitoring

Abstract

Health monitoring is an important category of mobile Health (mHealth) applications. Users generate a large volume of data during health monitoring, which can then be used by the mHealth server for constructing diagnosis or prognosis prediction models. However, these training samples contain private information of data owners, who may be reluctant to share them with the mHealth server. This paper proposes and experimentally studies a scheme that keeps the training samples private while enabling accurate construction of diagnosis and prognosis models. We specifically consider logistic regression models which are widely used in mHealth, and decompose the logistic regression model construction problem into small subproblems that can be executed by each user using their own private data. In this manner, users can keep their raw data locally and only upload encrypted parameters to the mHealth server for model construction. We show that our scheme suits well in mHealth applications by conducting experimental evaluations based on a real-world dataset and analyzing its computation overhead.